EEG correlates of conscious versus unconscious knowledge in artificial grammar learning

Abstract

The current study investigated the neural correlates of knowledge and the conscious status of knowledge in artificial grammar learning. In the training stage, subjects were exposed to grammatical strings which were displayed one letter a time. Subjects were asked to type the string back. In the test stage, subjects classified strings as to whether they had the same structure as the training strings. Then they indicated the basis of that judgment: guessing, intuition, familiarity, rules or memory. Tests strings violated repetition structure in the final (fifth) letter. In the test stage, the electroencephalogram was recorded while the fifth letter of each string was displayed. We were interested in two event‐related potential (ERP) contrasts: 1) grammatical strings versus matching ungrammatical strings which violated repetition structure; 2) unconscious structural knowledge attributions (random, intuition, and familiarity) versus conscious structural knowledge attributions (rules and recollection) when subjects classified correctly. Ungrammatical rather than grammatical strings elicited greater N2 amplitude at the frontal electrode positions (Fpz/Fz/F4/F6), indicating that N2 amplitude is a useful marker of knowledge of grammaticality. Additionally, the P300 component was higher for conscious rather than unconscious attributions at frontal‐central electrode positions (FCz/Cz C3/C5), indicating P300 is a useful marker of the conscious status of structural knowledge.